Research map / Label-efficient and robust learning / Recognition and long-tail
RGB-Infrared Cross-Modality Person Re-Identification via Joint Pixel and Feature Alignment
Guan'an Wang, Tianzhu Zhang, Jian Cheng, Si Liu, Yang Yang, Zengguang Hou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add pixel alignment module (ablation) | Basel. · SYSU-MM01 all-search single-shot | Rank-1 (%): 29.6 -> 40.6 | Improved |
| add feature alignment module (ablation) | Basel. · SYSU-MM01 all-search single-shot | Rank-1 (%): 29.6 -> 34.1 | Improved |
| add pixel alignment module (ablation) | Basel. · SYSU-MM01 all-search single-shot | mAP (%): 33.0 -> 38.7 | Improved |
| add feature alignment module (ablation) | Basel. · SYSU-MM01 all-search single-shot | mAP (%): 33.0 -> 36.2 | Improved |
| use pixel alignment instead of feature alignment (ablation) | FeatAlign · SYSU-MM01 all-search single-shot | Rank-1 (%): 34.1 -> 40.6 | Improved |
| stack pixel and feature alignment with two separate discriminators (ablation) | PixAlign · SYSU-MM01 all-search single-shot | Rank-1 (%): 40.6 -> 36.2 | Hurt |
| replace two separate discriminators with the joint discriminator (ablation) | AlignGAN- · SYSU-MM01 all-search single-shot | Rank-1 (%): 36.2 -> 42.4 | Improved |
| add joint discriminator on top of feature alignment only (ablation) | FeatAlign · SYSU-MM01 all-search single-shot | Rank-1 (%): 34.1 -> 42.4 | Improved |
| use AlignGAN instead of cmGAN | cmGAN · SYSU-MM01 all-search single-shot | Rank-1 (%): 27.0 -> 42.4 | Improved |
| use AlignGAN instead of cmGAN | cmGAN · SYSU-MM01 indoor-search single-shot | Rank-1 (%): 31.7 -> 45.9 | Improved |
| use AlignGAN instead of D-HSME | D-HSME · RegDB thermal2visible | Rank-1 (%): 50.2 -> 56.3 | Improved |
| use AlignGAN instead of D-HSME | D-HSME · RegDB visible2thermal | Rank-1 (%): 50.9 -> 57.9 | Improved |
Closest papers by meaning
- Cross-Modality Paired-Images Generation for RGB-Infrared Person Re-Identification (AAAI 2020 · similarity 0.94)
- RGB-Infrared Cross-Modality Person Re-Identification (ICCV 2017 · similarity 0.93)
- Neural Feature Search for RGB-Infrared Person Re-Identification (CVPR 2021 · similarity 0.90)
- Robust Multi-Modality Person Re-identification (AAAI 2021 · similarity 0.89)
- TokenMatcher: Diverse Tokens Matching for Unsupervised Visible-Infrared Person Re-Identification (AAAI 2025 · similarity 0.86)
- Cross-Modality Perturbation Synergy Attack for Person Re-identification (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
